Don’t use the ‘feed’ URI Scheme

Friday, 15 July 2005

It’s been covered before elsewhere, but just a friendly reminder: ‘feed’ URIs are bad for the Web, as are any that are used solely for dispatch (e.g., ‘itms’, ‘ pcast’).

I’m looking at you, Apple.

Interestingly, Apple also supports doing Web dispatch the right way; with media types. I know this because going to my test page — a zero-length file with a generic extension, but a ‘application/atom+xml’ media type — gets dispatched to my RSS reader in Safari on Tiger.

They also do content sniffing (the wrong way, but not so hideously evil and intrusive as creating a whole new URI scheme), because feeds that are served as text/plain ( for example) seem to get dispatched pretty reliably too.

So, why the new URI scheme? Complete cluelessness? A Pyrrhic plan to devalue the Web? What is it, Steve?

3 Comments

Randy Charles Morin said:

I just wanted to point out that USM [http://www.kbcafe.com/rss/usm.html] is and I hope you agree a better alternative for RSS. Actually, I’d like to know if you agree.

Friday, July 15 2005 at 11:10 AM

rjt said:

I’ve always wanted to know why OPML can’t be used for this.

Tuesday, July 19 2005 at 5:53 AM

Mark Nottingham said:

Randy,

I think you’re going in the right general direction; good stuff.

A couple of things; “application/rss+xml” isn’t registered, and isn’t likely to be, because the IESG wants what they consider a stable reference for the spec; e.g., an RFC. They don’t consider a URL good enough, so this is at a standstill. A more fruitful path would be to register a media type in the vendor or personal trees; e.g., application/prs.winer.rss+xml.

My inclination would be either to use a bare element or XLink for the RSS extension; that way, you don’t have to use Atom elements in RSS, which will confuse and/or turn off some people to this solution.

Wednesday, July 27 2005 at 6:22 AM

mark nottingham